Railroads and the Rural to Urban Transition: Evidence from 19th-Century Argentina∗ JOB MARKET PAPER Santiago P´erezy January 3, 2017 Download the latest version here Abstract I study the effects of improvements in transport infrastructure on the economic outcomes of parents and their children. To do so, I exploit the expansion of the railroad network in 19th-century Argentina and new longitudinal data following individuals before and after this expansion took place. To deal with the endogeneity of railroad location, I construct an in- strumental variable that takes advantage of the fact that districts along the route of province capitals were more likely to be connected. I find that, once their district got connected to the railroad, adults largely remained farmers or farm workers. By contrast, their children moved out of farming toward more modern and higher paying occupations. The movement out of farming occupations reflected both local changes in employment structure and increased migration out of rural areas, and it was more pronounced among children in districts where the soil was not suitable for agriculture. Consistent with the higher level of skills required for this transition out of farming occupations, children in connected districts were more likely to be literate in adult- hood. These results shed light on how improvements in transportation can shape the transition from a mostly rural to a diversified economy. ∗I thank my advisor Ran Abramitzky, as well as Arun Chandrasekhar, Victor Lavy, Melanie Morten and Gavin Wright for outstanding guidance as part of my dissertation committee. I am grateful to Raj Chetty, William Collins, Melissa Dell, Dave Donaldson, Pascaline Dupas, Marcel Fafchamps, James Feigenbaum, Joseph Ferrie, Price Fishback, Michela Giorcelli, Avner Greif, Caroline Hoxby, Claudia Rei, Mark Rosenzweig, Noam Yutchman and Ariell Zimran for detailed feedback. I am indebted to Stephen Redding and Pablo Fajgelbaum for sharing their GIS data on historical department borders of Argentina. I acknowledge the financial support of the Stanford University Economics Department, the Economic History Association, the Stanford Center for International Development and the Leonard W. Ely and Shirley R. Ely Graduate Student Fund Fellowship. yPh.D. Candidate, Stanford University, Department of Economics. email: [email protected], website: http://web.stanford.edu/ santip/ 1 Introduction Economists going back to North(1958) and Rostow(1959) have argued that improvements in transportation technology were a key driver of the transition from self-sufficient and mostly agrarian to modern economies. In this paper, I provide microeconomic evidence on the mechanisms through which reductions in transport costs can shape the shift of labor out of the agricultural sector. To do so, I exploit the expansion of the railroad network in 19th-century Argentina and new longitudinal data following individuals before and after this expansion took place. I show how households responded to this expansion through investments in human capital, occupational choices and migration decisions. To conduct the empirical analysis, I collect longitudinal data linking more than 30,000 males across the 1869 and 1895 national censuses of Argentina. In these data, I observe two groups of individuals. First, I observe children in the 1869 census and then link them to their long-term adult outcomes in 1895. Second, I observe working-age individuals in two points of their adult life. These data enable me to follow a large group of individuals as they transitioned across sectors of the economy and across places. To address the potentially endogenous placement of railroads, I use an instrumental variables approach. As a byproduct of connecting all province capitals, some intermediate districts not ex- plicitly targeted were nevertheless connected to the railroad network. My identification strategy exploits that, among these districts, those located along a convenient route { from a cost minimiz- ing perspective { were more likely to be connected. More precisely, I instrument railroad access with access to a hypothetical network that connects all province capitals to each other in a cost- minimizing way, a similar strategy to the one used by Chandra and Thompson(2000), Michaels (2008), Banerjee, Duflo, and Qian(2012), Faber(2014) and Morten and Oliveira(2015). By 1869, the year of the first national census, almost 70 percent of the population of Argentina resided in rural areas and was employed in the primary sector. Because of the absence of navigable rivers in the interior of the country, transportation was conducted using wagon carts that traveled at about 3 to 4 kilometers an hour (Conde, 1979). By 1895, when the second national census took place, the railroad network had expanded from less than 700 kilometers in 1869 to more than 14,000 kilometers and was the ninth longest network in the world. Railroads brought a dramatic decline 1 in the cost of moving goods and people. In 1874, traveling from Rosario to Cordoba (a 250-mile trip) took 15 hours by train { a trip that would have taken half a month in the pre-railroads era (Lewis, 1983). For most of the analysis, I focus on the probability of transitioning out of farming occupations as the main occupational outcome of interest. Understanding the drivers of this transition is important, as productivity in farming is typically lower than in the non-farming sector and this productivity gap is especially large in developing economies (Gollin et al., 2014; Vollrath, 2009). Railroad expansion could have influenced the exit out of agricultural occupations for three main reasons. First, railroads might have enabled individuals in rural areas to move physically. Railroads facilitated migration not only because they dramatically reduced travel times but also because they likely increased information flows across connected districts. Second, railroads might have allowed districts with relatively low productivity in farming to buy agricultural products from other districts. In this case, the decline in farming employment should have predominantly taken place in locations with low agricultural productivity. Finally, railroads could have also facilitated the adoption of labor-saving agricultural technology (Bustos, Caprettini, and Ponticelli, 2016). I first show that railroads had limited impacts on the probability that adults employed in farming or as farm workers in 1869 would transition out of these occupations. However, I find that railroads had large effects on the probability that the children of farmers and farm workers would exit these occupations. Children of farmers and farm workers in connected districts were approximately 13 percentage points more likely to work outside of farming occupations in adulthood, relative to a baseline probability of 32 percent. As they exited farming occupations, these children transitioned into white-collar and skilled blue-collar jobs in adulthood. How did railroads increase the probability of children exiting farming occupations? Children in connected districts might have transitioned out of farming by (1) exiting farming occupations but staying in their 1869 district of residence or (2) both exiting farming occupations and leaving their 1869 district of residence. I perform a simple decomposition exercise to estimate the relative contribution of these two factors. I find that 20 to 30 percent of the increase in the probability of exiting farming occupations can be attributed to the fact that children in connected districts were more likely to both migrate and transition out of farming occupations. This finding implies that railroads increased migration propensities and highlights the importance of tracking internal 2 migrants to obtain a comprehensive assessment of the impacts of transport infrastructure. At the same time, individuals who stayed in their 1869 district of residence accounted for an important fraction of the decline of farming in the children's generation. In particular, the shift away from farming was largely explained by children who in 1869 resided in districts where the soil was less suitable for agriculture. By contrast, there is limited evidence of a higher rate of exit out of farming occupations in connected districts with high agricultural suitability. This pattern is consistent with railroads reducing the overall demand for agricultural labor by facilitating specialization: railroads enabled districts where agriculture was relatively unproductive to buy agricultural goods from others. Why did the transition out of farming occupations occur in the children's generation but not among adults? The effects on adults might have differed from the effects on the children's generation for three primary reasons. First, adults who were employed in farming in 1869 had likely developed sector-specific skills, which made the transition out of this sector more costly for them than for their children (Matsuyama, 1992). Second, moving out of farming implied transitioning into occupations with higher skill requirements (Caselli and Coleman II, 2001), and adults who were already in the labor force by 1869 had likely fewer opportunities to acquire these skills than their children.1 Third, imperfections in the land market might have prevented individuals who worked their own land from transitioning out of farming. I find evidence that is more consistent with a mismatch between the skills of adults and the skills required in occupations outside farming than with frictions in the land market.